How good are you at Connect 4? Could you explain your strategy to someone else? Could you code it up as a computer program? The Arena is a web-based system for hosting strategy game tournaments, but where the matches are played by the computer programs people submit to the system. Join us and learn how to create and submit your own program, and take on the rest of the world at Connect 4!

How will you know if your vote is counted in the next election? With technological advances, and more and more elections being run on computers, trustworthiness and transparency are critical to public confidence. Join us to find out about the latest advances in electronic election systems, and how they can enable you, the voter, to check the integrity of an election.

A British Computer Society Lecture given by Dr Paul Schmidt, Portsmouth Hospitals NHS Trust

Information Security Research for the Move from "Need to Know" to "Need to Share" is part of a series of presentations during the semester.   Dr Adrian Waller is a Technical Consultant for Information Security at Thales Research and Technology (UK).

Selected undergraduate students from the Department of Computing at the University of Surrey will present the background and design stages of their final year project work. The event will provide the opportunity for the students to obtain feedback from a diverse audience of professionals, and will also allow the audience to discover the challenges facing current generations of near-graduate level students.

This year, we will hear about the Microsoft XNA computer game framework, an RSS tool that runs on the Google Android operating system, and a novel audio application.

This presentation, on security issues, is given by Chris Seary, Charteris plc.  Chris will be sharing his background both as a developer and as an auditor of large scale applications. 

The talk will focus on common pitfalls in development, touching on all aspects of the lifecycle, from design through to testing and deployment. Different technologies used within the Microsoft technology stack will be analysed, demonstrating how tools have progressed. 

Although Microsoft technologies are used for the demonstrations, much of the advice is applicable to other platforms. The presentation will also cover security solutions directly under the control of developers, such as WS-Security and fine-grain applications of encryption.

Chris Seary is an independent security consultant, providing advice to both Banking and Government. He is MVP, CISSP, ISO 27001 Lead Auditor, PCI DSS trained and CLAS. He frequently gives presentations and writes articles on IT Security. His specialism is securing enterprise scale applications.

This presentation introduces Mobile CSP||B, a formal framework based on CSP||B which enables us to specify and verify concurrent systems with mobile architecture as well as the previous static architecture. In Mobile CSP||B, a parallel combination of CSP processes act as the controller for the B machines and these B machines can be transferred between CSP processes during the system execution

Project Argus Professional is a multi-media Counter Terrorism presentation that takes the audience through an attack on a crowded city location. It analyses issues in the built environment that made the attack possible. It then challenges the audience to think how the likelihood of/or impact from such an attack could be reduced by the intelligent use of design and materials. The challenge for the 21st century is to make our built environment more resilient to terrorist attack without impinging on our ability to enjoy such places without draconian security measures.

This talk demonstrates the basic principles of biological evolution from the point of view of copying and mixing genetic data and goes on to illustrate how these principles have been used in evolutionary programming. Several published examples are examined in detail. Evolutionary programming is of interest for improving algorithms where the population of all possible algorithms is too large to be completely examined.

Paul King, Senior Security Adviser to CISCO, will give an overview of how organisations are at risk from corporate espionage - how organisations might be attacked and how they might reduce the risk. The talk will use Cisco's own organisation as an example. Paul will also discuss some of the research he is doing in this space.

A British Computer Society lecture:

Adrian's presentation will aim to answer these questions:
* What are some of the key cloud choice drivers?
* Can you identify primary transformational SHIFTS required to enable secure, but collaborative, clouds?
* Why does Identity and Access Management have to SHIFT?

Security is rarely seen as a business enabler - more an irksome and expensive disabler. Mike St John Green will construct the argument to show that it is an enabler, albeit expensive. We need to know how much to spend on security. He will show how to determine, in a systematic manner, the security features that are proportionate, answering the question, how much is good enough? Although Mike will be talking about how government tackles this problem, this issue applies to every business that relies on IT systems. This is really about risk management when applied to the security of IT systems.

Software reliability is a crucial factor to consider when developing software systems or defining optimal release time. For many organisations ‘time to market’ is critical and avoiding unnecessary testing time whilst retaining reliable software is important. 

From the recent Google Aurora attacks, to the 'dark market' organised crime networks, we are entering a new era of especially organised, motivated and sophisticated cyber-threats. It is therefore more critical than ever that businesses pro-actively manage the risks to their Information. 

Lecture by Dr. Bill Mitchell, Director of the BCS Academy

Computing education in schools is in a perilous state, university computing departments are under considerable strain and there is a pressing need for much better integration between the academic and business communities.

Computers and software have played a huge if relatively unexpected role in my engineering career, in my three businesses (and in most other people’s), in my early 1970’s rugby club and in my life. I first saw a computer 43 years ago, I first used one 41 years ago (and helped build them) and hardware still hates me! But software, engineering in the head, opened the doors for my future.
The question is, what is today opening the doors for your future?

Phil Smith will be discussing the tracking of surgical instruments for dexterity assessment with particle filters as his presentation for his MPhil-PhD transfer.

The style of medical training has emphasized more on standardized and objective assessment of clinical, academic and surgical knowledge. Traditionally in ophthalmology surgical skills are often assessed in the operating theatre environment with the supervising surgeon directly observing or providing feedback whilst watching a recording of the operation. This can be of great subjective variability and is not readily reproducible. Certain components of surgical skills can be determined by analyzing the movement of the instruments. 

Computer Scientists have a tendency to look at the systems they model and research from the angle of computer simulations. However, to clarify the nature of objects we simulate and mode, it is often worth to lean back and think a bit deeper out the nature of objects we are dealing with in a formal and mathematical framework since this can bring inconsistencies to light and give directions for new simulations.

Dr Runkler will give a short overview of the research activities on intelligent systems at Siemens Corporate Technology. In this talk, particular focus will be given to bio-inspired optimization methods and their applications, including ant colony optimization, wasp swarm optimization, fuzzy decision making and fuzzy weighted aggregation. These methods will be illustrated on several real-world industrial applications, including delivery logistics, cash management, car manufacturing, communication networks, maintenance scheduling, and electronics assembly.

The first in a series of NICE Research Group presentations.

We see the world around us in immense detail, in real time, and with no apparent effort. Yet we also seem to consistently misinterpret certain stimuli: optical illusions. 

In this presentation, Dr Corney will be demonstrating a variety of optical illusions and discussing their nature and cause. Dr Corney will present some recent work using artificial neural networks as a simple model of vision, including the perception of illusions, and he will also discuss the implications for other visual agents, including animals and robots

This is the next lecture in the series of NICE presentations.

Phacoemulsification is one of the core surgical skills in ophthalmic training and the most common procedure in ophthalmology. This work is a part of the project that aims to develop a tool to measure the surgical competence and technical skill through analysing the instrument movement in surgical videos. 

In this talk, Phil Smith will describe an approach that is able to track surgical instruments in cataract surgery using particle filters with a motion and colour based detector. Where experiments have shown it is possible to track an instrument even when prior information regarding its appearance is limited

As shared electronic data increases, it has become more difficult to manage it successfully and the demand for scalable and efficient mechanisms for managing and retrieving data effectively becomes essential. In this paper a more effective P2P architecture is presented, aiming to improve existing resource discovery processes. The proposed architecture is organised as a hierarchical super-peer structure, where super-peers of the network represent network’s knowledge that is formalised dynamically using its peers’ resources. The main focus of this paper is the creation of an adaptive hierarchical concept-based P2P topology using collective intelligence methods. In that process, unmanageable data is transformed into a structured knowledge based repository of semantic resources. Therefore, the network takes the form of an ontology of conceptually related entities of resource information, as provided by the peers. This knowledge driven approach has benefits over traditional load driven architectures, as the user query context is usually the main driver for managing the performance of the network, and in a way the network can be characterised as proactive rather than reactive. A number of experiments have been undertaken and results demonstrate the advantages of the proposed concept-based architecture over other popular architectures

Fraud is a serious and long term threat to a peaceful and democratic society; the total cost of fraud to the UK alone was estimated by the Association of Chief Police Officers to be at least £14bn a year. One such fraud is payment card fraud - to detect this fraud, organisations use a range of methods, with the majority employing some form of automated rules-based Fraud Management System (FMS). These rules are normally produced by experts and it is often an expensive and time-consuming task, requiring a high degree of skill. This analytical approach fails to address the fraud problem where the data and relationships change over time. 

Wireless Ad hoc Networks (WAHNs) offer a challenging environment for conventional Intrusion Detection Systems (IDSs). In particular WAHN have a dynamic topology, intermittent connectivity, resource constrained device nodes and possibly high node churn. Researchers over the past years have encouraged the use of agent-based IDS to overcome these challenges. In this work we propose the use of military tactics to optimise the operations of agent-based IDS for WAHN.

In this talk, Dr Lilian Tang will discuss the information variability in natural language and vision systems, the ambiguity caused by noisy data and their processing modules, and how context /reasoning can be modelled in order to perform application tasks in large-scale systems. This is to enable a system not just to deal with uncertainty and variability, but more to adapt to its unpredictable environment. She will review her progress so far in this form of contextual modelling and this will lead on to some open research challenges which now need to be addressed.

Until recently, the ever-increasing demand of computing power has been met on one hand by increasing the operating frequency of processors and on the other by designing architectures capable of exploiting parallelism at the instruction level through hardware mechanisms such as super-scalar execution. However, both these approaches seem to be reaching (or possibly have already reached) their practical limits, mainly due to issues related to design complexity and cost-effectiveness.

The current report explores the available supervised learning algorithms for multilayered spiking neural networks. Gradient descent based algorithms are one of the most used learning methods for rate neurons. The back-propagation version for spiking neurons firing a single spike, SpikeProp, promises the same learning abilities as for artificial neural networks. Systematic investigations on this learning method show that SpikeProp requires more computations than back-propagation and a reference start time is critical for convergence. These issues require significant improvements to the gradient descent learning method for spiking neural networks in order for an efficient algorithm to be developed. Further developments include a learning algorithm for input and output neurons with multiple spikes, and a general learning rule for recurrent networks.

Translational Research is one of the latest Government's key words; nowadays, if you want to get a grant then your scientific research needs to be translated into practical applications that have impact.  Working within a commercial environment is not straightforward but it is a two-way street, often the practical application helps to further drive the research. The application of neural computing to help industry address some pressing business needs has the potential to improve its performance in a number of keys areas. Over 15-years of delivering innovation into business, a series of projects and approaches are discussed.  Some have been successes, some have been failures, but what is the common theme?

Recently, much interest has developed in identifying reliable techniques that are capable of accurately ‘uncovering the truths’ regarding the pre and post- processing of a digital image, without the requirement of actively injecting a digital watermark or signature into the image data.  Whilst watermarking schemes have been shown to be useful for protecting the integrity of the image, there always exists the underlying risk that the watermark data might be forcibly or accidentally removed.  When this happens, the image is effectively stripped of its identity, and its integrity is extremely difficult to prove. Forensic techniques aspire to achieve similar objectives but do not rely on the strength of embedded data. Instead, the ambition is to identify the facts of an image, based solely on the data provided.

Wireless Ad hoc Networks (WAHN) offer a challenging environment for conventional intrusion detection systems (IDS). In particular WAHN have a dynamic topology, intermittent connectivity, resource constrained device nodes and possibly high node churn. Researchers over the past years have encouraged the use of agent-based IDS to overcome these challenges. 

The combination of a dominant (de-facto standard) design and associated search heuristics constitute a `technological paradigm'. Such technological paradigms may emerge as industries evolve, altering the nature of innovative search from exploration to incremental improvement along a `technological trajectory'. Disagreements exist as to the conditions of design standardisation and the relationship between standardisation and related shifts in innovation emphasis.

Jonathan Shaw from Lionhead Studios will be giving a talk in the normal CompSoc slot on Wednesday 1st December 2010 at 5:00pm in LTM (note the different venue for this week only).

Jonathan will be bringing along the Fable III SDK for a demo.  All are welcome.

The core concept of the proposed PhD research is to empower the accessibility of ICT and non-ICT technologies by introducing an innovative user modelling technique for the elderly and disabled. This new user modelling methodology will be able to describe in detail all the possible disabilities, the affected by the disabilities tasks as well as the physical, cognitive and behavioural/psychological characteristics of any user. An extension of UsiXML[1] language will be developed, in order to express the Virtual User Models in a machine-readable format. Research will be conducted in order to determine how the values of various disability parameters vary over individuals and if these values follow any common probability distribution (e.g.: Gaussian, Poisson, etc.). 

Most of the techniques and algorithms used for other applications such as communication, acoustics, and different biomedical engineering modalities can be extended to brain signal analysis. Spatial or temporal resolution limitation and the effect of noise and artifacts in the brain signals can be mitigated by processing of multichannel and/or multi-modal (such as joint EEG-fMRI) data using appropriate algorithms. Here, we may look at very recent techniques developed for analysis (noise and artifact removal, dynamics, source detection, localization, and tracking, prediction, etc.) of brain signals, and discuss various directions for future research.

Google CEO, Eric Schmidt, announced on the 15th of November 2010 the plan for their next generation of Android based mobile phones to become electronic wallets by making use of Near Field Technology (NFC). NFC is contactless technology based on high frequency RF tags already found in contactless cards like the Oyster. Little research has been carried out on how secure the services offered by NFC are, one of the reasons being its reliance on proximity (~10cm). Attacks that have been carried out used expensive antennas and other equipment. They have also been targeted at contactless cards and not mobile phones where other side channels exist like Bluetooth and Wi-Fi, making crosstalk and information leakage a security concern.

Humans communicate through different signals. Due to an ability to perceive things, our species can perform this communication accurately and efficiently even when noise is introduced or the signal is presented in different formats. Computer technologies aid in cognitive processing and can, to some degree, support intellectual performance and enrich individuals’ minds. We operate (and design systems that operate) using analogies, such as reasoning, comparisons, and synonymity. In the end: is analogy a shared abstraction? Does it derive from mathematics? Is it high-level perception in shared structure theory?

To build a scalable, robust and accurate P2P network, the network must be able to manage efficiently large amounts of information. Thus, a critical challenge in P2P networks is to collectively transform resources to a repository of semantic knowledge to  accurately and efficiently discover resources. 

In everyday life human decision-making is often based on arguments and counter-arguments. Decisions made in this way have a basis that can be easily referred to for explanation purposes as not only is a best choice suggested, but also the reasons of this recommendation can be provided in a format that is easy to grasp.

Much research work has been done on formalizing UML diagrams, but less has focused on using this formalization to analyze the dynamic behaviours between formalized components. In this work we propose using a subset of fUML (Foundational Subset for Executable UML) as a semi-formal language, and formalizing it to the process algebraic specification language CSP, to make use of FDR2 as a model checker.

The hippocampus is widely believed to mediate mammalian declarative memory function, and it has been demonstrated that single pyramidal neurons in this cortical region can encode for the presence of multiple spatial and non-spatial stimuli. Furthermore, the rate and phase of firing - with respect to theta oscillations in the local field potential – can be dissociated, and may thus encode for separate variables. This has led to the suggestion that hippocampal processing may operate using a dual (rate and temporal) coding mechanism.

This PhD project aims to realize a new type of information system, more dynamic and less opaque to its owners, specified with structured natural language models and queried through hypermedia. To accomplish this, we focus on Semantics of Business Vocabulary and Rules (SBVR) as a modelling language, Representational State Transfer (REST) as an interface paradigm and Relational Databases as the persistence mechanism. All three of these technologies have declarative underpinnings, focusing on the ‘what’ rather than the ‘how’, which is why their combination is feasible and effective. By creating appropriate mappings to align these technologies, we create a core platform for Generative Information Systems.

Most of the techniques and algorithms used for other applications such as communication, acoustics, and different biomedical engineering modalities can be extended to brain signal analysis. Spatial or temporal resolution limitation and the effect of noise and artifacts in the brain signals can be mitigated by processing of multichannel and/or multi-modal (such as joint EEG-fMRI) data using appropriate algorithms. 

Here, we may look at very recent techniques developed for analysis (noise and artifact removal, dynamics, source detection, localization, and tracking, prediction, etc.) of brain signals, and discuss various directions for future research.

The nervous system is under selective pressure to generate adaptive behavior but at the same time is subject to costs related to the amount of energy neural signalling consumes. Characterizing this cost-benefit trade-off is essential for understanding the function and evolution of nervous systems, including our own. 

FATIGUE is a common phenomenon that exists in our everyday life which is the state of reduced performance and can have mental or physical component. 

The state of reduced performance of the operators that relates to the fatigue has been caused many disasters which many of them are not well known to the public.

A scheme is presented in which a Pretty Good Democracy style acknowledgement code mechanism is incorporated into Prêt a Voter. Voters get immediate confirmation at the time of casting of the correct registration of their receipt on the Web Bulletin Board. As with PGD, the registration and revealing of the acknowledgement code is performed by a threshold set of Trustees. Verification of the registration of the vote is now part of the vote casting and therefore more convenient for the voters. This verification mechanism supplements the usual verification on the Web Bulletin Board mechanism, that is still available to voters. 

MLS has been a field of study in computer science for decades, and MLS systems have been developed and deployed for high assurance defence and government applications. However, in recent years other users with less stringent security requirements have been talking about their need for "MLS", and have been attempting to use traditional MLS solutions in their systems. In this talk, we take a look at the varied applications that are claimed to require "MLS" and attempt to reconcile their different interpretations of the term. We then survey existing and proposed MLS technologies, discuss some of their drawbacks when compared with these applications' requirements, and propose some areas for future research.

The speaker is Dr Adrian Waller, Technical Consultant at Thales.

For the H.264/AVC video stream with covert data by collusion-irresistent steganography, a blind video steganalysis scheme is proposed based on the collusion sensitivity and noise classification. It exploits temporal frames weighted averaging (TFWA) for collusion to improve the capabilities of host approximation and watermark removal, instead of the traditional temporal frames averaging (TFA). To overcome the interferences by motion and illuminance variation, a content change factor (CCF) is defined to adaptively classify the noise existing in prediction error frames (PEF). For passive steganalysis, final decision is made by the center of mass (COM) feature of histogram characteristic function (HCF). Experimental results demonstrate that the proposed approach can cope with temporal-domain, transform-domain and even spread spectrum based steganographic algorithms. For a stego-video with embedding strength of 10%, it can achieve a probability of correct detection (PCD) about 99.82%.

Compressed Sensing (CS) framework which is linked with the sparse recovery problem has been recently introduced and applied to solve numerous problems.  Measurement matrix has a key role in the CS to sample the signal/images. It has been recently shown that optimization of this matrix can increase the quality of reconstruction.  In this talk we first introduce the CS theory.  Then, the advantages of the measurement matrix optimization and our proposed strategies for this purpose are discussed.  

Finally, we review some applications and extensions of CS such as functional Magnetic Resonance Imaging (fMRI) and Watermarking.

Non-negative matrix factorization (NMF) has been widely used for analyzing multivariate data. NMF is a method which creates a low rank approximation for positive data matrix and because of non-negativity constraint it has found interesting applications in image processing where he data is inherently positive.  Functional Magnetic Resonance Imaging (fMRI) is an imaging technique which provides useful anatomical and functional information of brain.  Analyzing data provided by the fMRI helps to investigate brain function.  

In this talk, we first give a brief introduction about different algorithms for fMRI analysis. Then, the application of Non-negative matrix factorization to fMRI data and our proposed algorithm for this purpose will be discussed and its superiority to other data decomposition techniques such as BSS will be emphasised for such data.

Neural networks based on temporal encoding with single spikes are biologically more realistic models as experimental evidence suggests that biological neural systems use the exact time of action potentials to encode information.  Moreover, it has been demonstrated that networks of spiking neurons are computationally more powerful than sigmoidal neurons.  In order to reach the computational power of spiking neurons, efficient learning algorithms must be used.  This presentation explores the available supervised learning methods in artificial and spiking neural networks

Professor Seth Grant from the Genes to Cognition Programme, Wellcome Trust Sanger Institute, Hinxton, Cambridge, will visit the University of Surrey to give a presentation to the Department of Computing and all are welcome to attend.  

For over a century it has been known that the synapse – the junction between nerve cells – is of fundamental importance in organizing brain circuits and behavior.  

Near field communication (NFC) is a standard-based wireless communication technology that allows data to be exchanged between devices that are a few centimeters apart. In this presentation three major elements will be discussed: the concept of NFC, the different protocols used (their pros and cons) and the potential areas of research.

A British Computer Society event.  This event is open to Members and Non-Members. Students are particularly welcome.

Please see the Branch website for further details.

http://www.guildford.bcs.org

Democracy entirely depends on the elections, which must be robust and fair without cheating and electoral frauds. Voters must be sure that their vote has remained unaltered and has been correctly tallied. The election system should prevent any possible coercion and should be robust even if the official authorities are not trusted. There is a recent example where frauds and systems’ misbehaviour were reported by voters (Florida, 2000). When security properties like integrity, privacy, anonymity, confidentiality and verifiability are not supported or they have limited functionalities, attacks can be made enabling a third party to learn the voters vote. All these lead to a fundamental question: how can the voter trust the voting procedure and the announced results?

A long standing aim for computer science has been to build ‘intelligent machines’. Building computer models of the human brain has the potential to achieve this aim, either through developing an artificial brain, or by understanding how the brain computes to replicate intelligence. However, despite significant advances, we have yet to realise this potential.  There appears to be a clear gap between modelling the brain for neuroscience and applying what we have learnt about brain-like computation to real-world, practical problems. On the one hand, models based on high-level cognition have been developed which can process real-world inputs. These cognitive architectures may show us broadly how the brain achieves certain function, but are too simplistic for practical purposes. On the other hand, large-scale brain simulations have been developed which model brain dynamics, but are not designed to replicate intelligence.  In this seminar we will explore these issues and debate some of the possible long-term answers which involve bridging the gap between cognitive architectures and large-scale simulations, particularly for sensory processing.

From the recent Google Aurora attacks, to the 'dark market' organised crime networks, we are entering a new era of especially organised, motivated and sophisticated cyber-threats.  It is therefore more critical than ever that businesses pro-actively manage the risks to their information.

It is well established that the phenomena of synaptic, intrinsic and homeostatic plasticity are mediated – at least in part – by a multitude of activity-dependent gene transcription and translation processes.  Various isolated aspects of the complex genetic regulatory network (GRN) underlying these interconnected plasticity mechanisms have been examined previously in detailed computational models.  However, no study has yet taken an integrated, systems biology approach to examining the emergent dynamics of these interacting elements over longer timescales.  Here, we present theoretical descriptions and kinetic models of the principle mechanisms responsible for synaptic and neuronal plasticity within a single simulated Hodgkin-Huxley neuron.  We describe how intracellular Calcium dynamics and neural activity mediate synaptic tagging and capture (STC), bistable CaMKII auto-phosphorylation, nuclear CREB activation via multiple converging secondary messenger pathways, and the activity-dependent accumulation of immediate early genes (IEGs) controlling homeostatic plasticity.  We then demonstrate that this unified model allows a wide range of experimental plasticity data to be replicated.  Furthermore, we describe how this model can be used to examine the cell-wide and synapse-specific effects of various activity regimes and putative pharmacological manipulations on neural processing over short and long timescales.  These include an examination of the interaction between intrinsic and synaptic plasticity, each dictated by the level of activated CREB; and the differences in functionality generated by STC under regimes of reduced protein synthesis.  Finally, we discuss how these processes might contribute to maintaining an appropriate regime for transient dynamics in putative cell assemblies within contemporary neural network models of cognitive processing.

Chris Seary, Consultant at Charteris, will be giving two talks. The first will cover real world application security from an auditor's perspective. It goes through many of the common security issues arising from lack of secure development practice. This will give demonstrations of injection attacks on a web site.

The second talk will cover the newer WS-Security toolset for SOAP web services. It will show examples of code, configuration and the communications used.

The algorithm takes full advantage of local correlation of the host image pixels and the masking characteristics of the human visual system, it chooses DCT blocks by comparing the value of the DCT low frequency coefficients and the amount of the nonzero DCT coefficients of each block. After the embedding process is completed, transforming the DCT coefficients from the frequency domain to the spatial domain produces some rounding errors, because the conversion of real numbers to integers will cause some information loss. The paper uses genetic algorithm to deal with the rounding errors. The experimental results show that, the algorithm can not only make sure the quality of the embedded image and the invisibility of the watermark, but also robust to common image operates, and JPEG compress

Machine learning is inherently a multi-objective task. Traditionally, however, either only one of the objectives is adopted as the cost function or multiple objectives are aggregated to a scalar cost function. This can be mainly attributed to the fact that most conventional learning algorithms can only deal with a scalar cost function. Over the last decade, efforts on solving machine learning problems using the Pareto-based multi-objective optimization methodology have gained increasing impetus, thanks to the great success in multi-objective optimization using evolutionary algorithms and other population-based stochastic search methods.

Compared with Generalized Gaussian distribution (GGD), Cauchy distribution is superior to describe the statistical distribution of the Intra-coded DCT coefficients in H.264/AVC. For the bipolar additive watermark in H.264/AVC video stream, a Cauchy distribution based detection algorithm is proposed by ternary hypothesis testing. Experimental results show that the proposed approach can achieve more than 80% on average for the accuracy of watermark detection.

Professor Chris Mitchell, from Royal Holloway, will be our next speaker.  

This cautionary tale is about an apparently trivial software project which didn't go too well. After posing some questions, it explains what happened, mentions some useful survival tools and techniques that would certainly have made things go better and ends with a couple of stories with enduring relevance. Relevant are 'luck' and quality – corner stones of personal success which are hard to define and even harder to achieve.

Programming is a major subject in Computer Science (CS) departments.  However, students often face difficulties on the basic programming courses due to several factors that cause these difficulties.  Maybe the most important reason is the lack of problem solving abilities that many students show.  Due to their lack of general problem solving abilities, students don’t know how to create algorithms, resulting in them not knowing how to program.

Working memory refers to a limited capacity part of the human memory system that is responsible for the temporary storage and processing of information while cognitive tasks are performed.  We will explore how memories can be represented by extensively overlapping groups of neurons that exhibit stereotypical time-locked spatiotemporal time-spiking patterns, called polychronous patterns and make further assumptions regarding the polychronous group span of different brain regions in association to working memory.

Cyberwarfare is a subject that has received a great deal publicity since the attacks on Estonia and Georgia and the Stuxnet malware in particular. Nation states are now devoting much more attention to the damage that can be inflicted upon them without a shot being fired and the probability that it will happen to them.

This research proposes a method for developing a data warehouse in a clinical environment while particularly focusing on the requirements specification phase. It is conducted primarily to target organizations whose requirements are not clearly defined and are not yet aware of the benefits of implementing a data warehouse. By integrating key ideas such as the agile manifesto, maintaining data quality, and incremental and prototyping approaches, it provides a platform for collaboration and participation between users and designers, as well as identifying relevant processes and their additional value. It is also important to note that this work was performed in the context of a Clinical Unit with limited experience of IT, and limited budget. An important research objective was to demonstrate how to obtain significant “buy-in” to a data warehouse solution at low-cost, and minimal risk to the clinical unit.

Fair elections are essential processes in ideal representative democracies since ancient Greece.  Thus, as being an indispensable part of fair elections, a various number of trustworthy voting systems has been designed and improved over decades.  However, due to insufficient amount of proofs, the lack of trustworthiness of such systems still precipitates quite a number of system attacks violating citizens' privacy, modifying election results, which have as a consequence controversial elections and unfair democracies.

Data Mining is the use of algorithms to extract the information and patterns derived by the knowledge discovery in databases process.  Classification is a major data mining task.

Classification maps data into predefined groups or classes. It is often referred to as supervised learning because the classes are determined before examining the data. In this research work, new hybrid classification methods are proposed using classifiers in a heterogeneous environment with using voting and stacking mechanisms and their performances are analyzed in terms of error rate and accuracy.

A Classifier ensemble was designed using a k-Nearest Neighbour (k-NN), Radial Basis Function (RBF), Multilayer Perceptron (MLP), and Support Vector Machine (SVM). The feasibility and the benefits of the proposed approaches are demonstrated by means of data sets like intrusion detection in computer networks, direct marketing, signature verification. Experimental results demonstrate that the proposed hybrid methods provide significant improvement of prediction accuracy compared to individual classifiers.

Dr Kianoush Nazarpour from the Motor Control Group at the Institute of Neuroscience, Newcastle University will be our speaker on this occasion.  All are welcome to attend.

Biomimetic brain-machine interfaces (BMI) have evolved from experimental paradigms exploring the neural coding of natural arm and hand movements to mathematically advanced real-time neural firing rates decoders.  However, despite recent decoding algorithms with increasing levels of performance and sophistication, BMI control remains slow and clumsy in comparison to natural movements.  Therefore, considerable improvements are required if these devices are to have real-life clinical applications.

This work suggests that population-level selection of artefact designs produced by firms facing an ill-structured design problem favours the formation of a dominant design with a set of 'high pleiotropy' elements, affecting many product functions. Selective expansion of a technological artefact's active 'design space' may embed a negative heuristic within the design, effectively 'locking-in' earlier design choices. In the absence of sufficient variety-generating mechanisms, competition will result in a dominant design within an industry. This research describes how selection at the population level may interact with the local search routines of firms to produce a dominant design embodying 'frozen' dimensions. Such a design may be seen to form part of a technological paradigm. The investigation nests Koen Frenken's existing model of technological paradigms within an evolutionary population-based model. Entropy statistics indicate several exploratory stages that emerge endogenously via interaction of selection at the firm and population levels

 Our speaker will be Dr Markus Roggenbach, from the Department of Computer Science, Swansea University.

Taking the "Children & Candy Puzzle" (see below) as a master example, we discuss what the various tools for the process algebra CSP offer for modelling and verifying systems.  Here, we focus especially on interactive theorem proving for CSP, as exemplified in Steve Schneider's work or in our tool CSP-Prover.  Besides the power to analyze infinite state systems, the theorem proving approach offers the possibility for deeper reflections on CSP.  Here we discuss how it allows one to verify the algebraic laws of the language, and, furthermore, how it allows to prove meta results such as the completeness of axiomatic semantics.

Children & Candy Puzzle: "There are k children sitting in a circle.  In the beginning, each child holds an even number of candies.  The following step is repeated indefinitely:  Every child passes half of her candies to the child on her left; any child who is left with an odd number of candies is given another candy from the teacher.  Claim: Eventually, all children will hold the same number of candies."  

Bayesian Networks (BN) are an increasingly important tool for mining complex relations in large data sets.  A major focus of current research is in efficient and effective ways of learning those essential interactions between variables, known as structure, that allow efficient factorisation of the joint probability distribution of the data.  This in turn provides a platform for prediction, inference and simulation.

In this research, we propose a camera identification technique based on the conditional probability features (CP features). Specifically we focus on its performance for detection of images sources which has been taken using cameras from different models. By using 4 cameras, we demonstrate that the CP features are able to perfectly match the test images with its source in 8 over 10 independent tests conducted. Additionally, the CP features are also able to perfectly match the cropped and compressed test images with its source in 9 over 10 independent tests. These findings provide a good indication that CP features are beneficial in image forensics.

The primary benefit of digital content, the ease with which it can be duplicated and disseminated, is also the primary concern when endeavouring to protect the rights of those creating the content. Copyright owners wish to deter illicit file sharing of copyrighted material, detect it when it occurs and even trace the original perpetrator. Embedding a unique identifying watermark into licensed multimedia content enables those selling digital content to trace illicit acts of file sharing to a single transaction with a single a buyer. However, evidence of such illicit activity must be gathered if and only if the buyer truly shared the content for a seller to prove such behaviour to an arbitrator. For this purpose, Buyer-Seller Watermarking (BSW) protocols have been developed to be used in conjunction with digital watermarking schemes.

Since the introduction of Huffman codes back in 1952, variable-length codes have been used in several data compression standards, including the latest video coding standards such as H.264, usually as part of their entropy coding sub-systems.  Although not as good as other data compression schemes, such as for example Arithmetic Coding, they still prove popular in practical implementations due to their simplicity.  However, from early on it was realized that variable-length codes suffer from error-propagation under noisy conditions. Several techniques have been proposed to mitigate this behaviour, including the use of synchronisation codewords, self-synchronising codes and reversible variable-length codes.

In this talk, Dr Shan He will briefly introduce his multi-disciplinary research in the areas of computational intelligence and computational biology. 

Firstly he will introduce his ongoing research in applying Computational Intelligence, e.g., evolutionary computation to metabolomics.  Then Dr He will present a novel ensemble-based feature selection algorithm for discovering putative biomarkers from high-dimensional omics data. Finally, he will present his work in simulating the evolution of animal self-organising behaviour using evolutionary agent-based modelling.

We investigate interactions between evolution, development and lifelong layered learning in a combination we call Evolutionary Developmental Evaluation (EDE).  It is based on a specific implementation,  Developmental Tree-Adjoining Grammar Guided GP (DTAG3P).

To date, diffusion tensor imaging (DTI) is the only non-invasive tool available to reveal the neural architecture of human brain white matter.  Advances in DTI techniques have shown great potential in the study of brain white matter related diseases such as depression, traumatic brain injury and Alzheimer's disease (AD).  In DTI, a reliable reconstruction of neural fibre structure relies on the accurate estimation of fibre orientation distribution function (fODF) from each individual voxel in diffusion weighted images (DWI).

With the tremendous growth and usage of digital images nowadays, the integrity and authenticity of digital content is becoming increasingly important and of major concern to many government and commercial sectors. In the past decade or so, digital watermarking has attracted much attention and offers some real solutions in protecting the copyright and authenticating the digital images. Four novel robust and semi-fragile transform based image watermarking related schemes are introduced. These include wavelet-based contourlet transform (WBCT) for both robust and semi-fragile watermarking, slant transform (SLT) for semi-fragile watermarking as well as applying the generalised Benford’s Law to estimate JPEG compression, then adjust the appropriate threshold for improving the semi-fragile watermarking technique.

Video Watermarking and Forensics is the next talk in the Watermarking series of events.

Advances in electronic voting have made it possible to run more robust and transparent elections than previously possible with traditional paper based voting. The higher security guarantees given through electronic voting intends to raise standards of democracy in modern society and reduce the possibility of election malpractices. Despite having these obvious benefits, the rate at which electronic voting systems are adopted across the world, especially for legally binding elections, has been very slow. Only a few countries have successfully conducted elections at national level such as Estonia and Switzerland. The voter population in these countries are comparatively lower and the voting systems are in most instances integrated to existing security infrastructures.  However, in recent elections there has been speculation as to the adequacy of security used for these elections.

Monitoring mental fatigue is a crucial and important step for prevention of fatal accidents.  This may be achieved by understanding and analysis of brain electrical potentials.  Electroencephalography (EEG) is the record of electrical activity of the brain and gives the possibility of studying brain functionality with a high temporal resolution.  EEG has been used as an important tool by researchers for detection of fatigue state.  However, their proposed methods have been limited to classical statistical solutions and the results given by different researchers are somehow conflicting.  Therefore, there is a need for modification of the existing methods for reliable analysis of mental fatigue and detection of fatigue state. 

Recent technological advances have caused an infrastructural paradigm shift and the rapid growth of communities that are connected by virtual means.  The value of the Web is growing constantly, with ever more users joining and contributing to the network.  Fortunately, unlike conventional social networks, the connections in a virtual setting are clearly visible for analysis

The orthogonal least squares (OLS) algorithm, developed in the late 1980s for nonlinear system modelling, remains highly popular for nonlinear data modelling practicians, for the reason that the algorithm is simple and efficient, and is capable of producing parsimonious nonlinear models with good generalisation performance.  Since its derivation, many enhanced variants of the OLS forward regression have been developed by incorporating the recent developments from machine learning.  Notably, regularisation techniques, optimal experimental design methods and leave-one-out cross validation have been combined with the OLS algorithm.  The resultant class of OLS algorithms offers the state-of-the-arts for parsimonious modelling from large data.  

Our next Department Seminar speaker will be Dr Alastair MacWillson, Global Managing Director at Accenture Technology Consulting, London.  

With the number of ‘cyber attacks’ on the rise, targeting government and industry across the globe, it is clear that most organisations are now facing a whole new category of threat.  At the same time, inherent weaknesses in enterprise IT and ineffective approaches to information security are putting organisations at risk as never before.   There is a growing realisation that confronting these advanced threats calls for a whole new doctrine of defence.  Keeping pace with the digital arms race requires constantly re-evaluating your position against the threats and adapting your information security strategies.   Intelligence gathering has become an essential core competency for every security team.   

Steganography is the art of providing a secret communication channel for the transmission of covert information. At the same time, it is possible that it can be used by cybercriminals to conceal their works. This potential illegal use of steganography is the basis for the objectives in this thesis. This thesis initially reviews the possible flaws in current implementations of steganalysis. By using images from different camera types, this thesis confirms the expectation that the steganalysis performance is significantly affected by the differences in image sources. In this thesis we prove that image compression in a steganalysis process has an impact on the steganalysis performance, as claimed in the literature. A review of currently available steganalysis techniques, along with a proposal to overcome the said problems is also presented in this thesis. 

The autumn series of Nature Inspired Computing and Engineering Research Group seminars begins with our first seminar on Thursday 6th October. 

This talk will present a controller tuning algorithm inspired by the ‘Bayesian brain’ hypothesis; that is the brain models its environment in terms of probabilities, and uses approaches similar to those used in Bayesian statistics to make decisions. The tuning algorithm combines this theory with current understanding of neuromodulatory system, specifically the idea that neuromodulation is a mechanism for adjusting the hyperparameters of learning algorithms. It has been applied to three different components of a walking robot controller; the leg coordination component, which guides the robot towards a target while avoiding obstacles, the trajectory planning component which calculates the paths of each individual leg, and the tracking controller, which ensures the desired path is followed. The final controller demonstrates adaptability and robustness, as well as being reliable and improving efficiency by reducing power and torque requirements..

This seminar given by Dr Shujun Li will kick off a series of seminars run by the MSF group.

A general method for recovering missing DCT coefficients in DCT-transformed images is presented in this work. We model the DCT coefficients recovery problem as an optimization problem and recover all missing DCT coefficients via linear programming. The visual quality of the recovered image gradually decreases as the number of missing DCT coefficients increases. For some images, the quality is surprisingly good even when more than 10 most significant DCT coefficients are missing. When only the DC coefficient is missing, the proposed algorithm outperforms existing methods according to experimental results conducted on 200 test images. The proposed recovery method can be used for cryptanalysis of DCT based selective encryption schemes and other applications. We also discuss possible extension of the optimization model to some other problems in multimedia coding, and security and forensics.

Most of the subspace based signal separation methods are applicable when sufficient number of signal mixtures are available which require multichannel recordings. Separation of signal sources from single channel recordings on the other hand is often of very poor quality, if not impossible, since so-called subspaces of the signal components are unknown. 

Singular spectrum analysis (SSA) deals with decomposition of the data into more meaningful subspaces where the desired signal components are characterised. Periodic signals, spikes, and those for which some a priori knowledge is available can be well defined in the eigenspace of the SSA. Some applications of this approach will be explained and a new SSA-based adaptive filter for recovery of periodic physiological signals from their single channel mixtures will be presented.

Next in the series of the autumn seminars.  All are welcome to attend.  Details to follow.

A British Computer Society event.  This event is open to Members and Non-Members. Students are particularly welcome.

Please see the Branch website for further details.

http://www.guildford.bcs.org